Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 41 of the 41 references it cites were found in a registry or an index.

Recurrent neural networks in synthetic cells: a route to autonomous molecular agents?

Open access
Braccini, Michele, Collinson, Ethan, Roli, Andrea, Fellermann, Harold, Stano, Pasquale · 2023 · Frontiers in Bioengineering and Biotechnology

Abstract

Prompted by recent advancements in synthetic biology, we highlight the possible use of bio-chemical reactivity to generate tools and strategies for a genuinely new AI in the wetware domain. Chemical neural networks (CNNs) could be realized inside synthetic cells by employing elements of bacterial two-component signaling systems. The major novelty of CNNs, when compared to neural networks, consists in the embodiment of the network nodes and links: these network elements are no more logical entities but physical ones, whose behavior is subjected to the physico-chemical laws. Moreover, the results of network computation (i.e., molecules) belong to the same domain as the network elements, the physical domain. Building on this, we illustrate a viable way for implementing recurrent links in CNNs, making them able to have an internal state, i.e. a memory. This capability makes it possible to endow CNNs with some sort of autonomy, as they can take decisions also on the basis of their state and not just as a consequence of the computation of the inputs. At a more fundamental level, synthetic cell technology can be a platform for crucial investigations of theoretical biology principles. For instance, autonomy can be seen as a prerequisite for agency and other more complex characteristics of living beings.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3389/fbioe.2023.1210334

Cites

41 of 41 references verified0 citation errors · 0 not checkable

Statistics

Figures

FIGURE 1 Recurrent Neural Networks in Synthetic Cells. (A) Synthetic/artificial cells (SCs) can be constructed via the so-called “bottom-up” synthetic biology approach by several methods that lead to the guided assembly of selected molecular components such as lipids, DNA, ribosomes, enzymes, tRNAs, small molecules, natural or artificial organellae into cell-like structures that roughly resemble the structure and the function of living biological cells. To date, several non-trivial SCs have been built in the lab, ranging from SCs hosting one or more enzyme-catalized pathways, gene expression, lipid synthesis, chemical signaling, etc. (B) Typical appearance of SCs (by confocal fluorescence microscopy) producing the green fluorescent protein in their inner volume. The membrane is stained by including a red-fluorescent marker. The size bar represents 25 μm. (C) Schematic representation of two-component signaling systems, which enable bacteria to sense, respond, and adapt to their environments, by letting the cell perceive chemical signals present in their surroundings. In a typical system, a membrane protein (sensor) with histidine kinase activity catalyzes its auto-phosphorylation in the presence of an extracellular stimulus x. Next, the sensor is capable of transferring the phosphoryl group to a response regulator, which–thanks to this activation–can then affect cellular physiology by regulating gene expression or by modulating protein activity. (D) The mentioned two-component signaling systems can cross-talk (or can be engineered, in principle, in order to enhance cross-talk) so to realize a sort of chemical neural network based on the phosphorylation cascades (called “phospho-neural networks” by (Hellingwerf et al., 1995), as discussed by (Gentili and Stano, 2022; Stano et al., 2022)). In particular, sensors S1 and S2, response regulator R1 and R2, and genes G1 and G2 realize a small chemical neural network with [S1, S2] as input layer, [R1, R2] as hidden layer, and [G1, G2] as output layer. The network performs the computation of extracellular signals (x,y) into intracellular effects (p,q). For example, x and y are small molecules and p and q are proteins affecting the cellular state. As evidenced by the bottom diagram, the time evolution of the states S of the cell (intended as an “agent”) depends on the states E of the environment. (E) The network drawn in (D) can be transformed in a recurrent chemical neural network if at least one of the outputs is allowed to affect the computation carried out by one of the nodes [S1, S2, R1, R2]. It is possible to imagine several ways this can happen (increase or decrease of sensor and/or response regulator concentration, allosteric regulation by a third-party component). Now, and in contrast with panel (D), the state S of the agent (e.g., S t+1) will depend not only on the state E t of the environment, but also on the state S t of the agent. In other words, the agent state co-determines, with E, the next agent state. The relative strengths of these two dependencies (e.g., the “weights” of the arrows pointing from E t and S t to S t+1) will measure the degree of autonomy of the network (and of the agent). The recurrent CNN can be interpreted as a control module that confers the SCs in which it is embedded a certain degree of autonomy. The bottom diagrams shown in (D,E) have been adapted from (Bertschinger et al., 2008).
FIGURE 1 Recurrent Neural Networks in Synthetic Cells. (A) Synthetic/artificial cells (SCs) can be constructed via the so-called “bottom-up” synthetic biology approach by several methods that lead to the guided assembly of selected molecular components such as lipids, DNA, ribosomes, enzymes, tRNAs, small molecules, natural or artificial organellae into cell-like structures that roughly resemble the structure and the function of living biological cells. To date, several non-trivial SCs have been built in the lab, ranging from SCs hosting one or more enzyme-catalized pathways, gene expression, lipid synthesis, chemical signaling, etc. (B) Typical appearance of SCs (by confocal fluorescence microscopy) producing the green fluorescent protein in their inner volume. The membrane is stained by including a red-fluorescent marker. The size bar represents 25 μm. (C) Schematic representation of two-component signaling systems, which enable bacteria to sense, respond, and adapt to their environments, by letting the cell perceive chemical signals present in their surroundings. In a typical system, a membrane protein (sensor) with histidine kinase activity catalyzes its auto-phosphorylation in the presence of an extracellular stimulus x. Next, the sensor is capable of transferring the phosphoryl group to a response regulator, which–thanks to this activation–can then affect cellular physiology by regulating gene expression or by modulating protein activity. (D) The mentioned two-component signaling systems can cross-talk (or can be engineered, in principle, in order to enhance cross-talk) so to realize a sort of chemical neural network based on the phosphorylation cascades (called “phospho-neural networks” by (Hellingwerf et al., 1995), as discussed by (Gentili and Stano, 2022; Stano et al., 2022)). In particular, sensors S1 and S2, response regulator R1 and R2, and genes G1 and G2 realize a small chemical neural network with [S1, S2] as input layer, [R1, R2] as hidden layer, and [G1, G2] as output layer. The network performs the computation of extracellular signals (x,y) into intracellular effects (p,q). For example, x and y are small molecules and p and q are proteins affecting the cellular state. As evidenced by the bottom diagram, the time evolution of the states S of the cell (intended as an “agent”) depends on the states E of the environment. (E) The network drawn in (D) can be transformed in a recurrent chemical neural network if at least one of the outputs is allowed to affect the computation carried out by one of the nodes [S1, S2, R1, R2]. It is possible to imagine several ways this can happen (increase or decrease of sensor and/or response regulator concentration, allosteric regulation by a third-party component). Now, and in contrast with panel (D), the state S of the agent (e.g., S t+1) will depend not only on the state E t of the environment, but also on the state S t of the agent. In other words, the agent state co-determines, with E, the next agent state. The relative strengths of these two dependencies (e.g., the “weights” of the arrows pointing from E t and S t to S t+1) will measure the degree of autonomy of the network (and of the agent). The recurrent CNN can be interpreted as a control module that confers the SCs in which it is embedded a certain degree of autonomy. The bottom diagrams shown in (D,E) have been adapted from (Bertschinger et al., 2008).CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 41 references

Research works printed but not linked: 1

  1. Langton, 1989, Artificial life, 1

Cited by

See all 18

Metadata

Authors
Braccini, M., Collinson, E., Roli, A., Fellermann, H., & Stano, P.
Affiliations
  • University of Bologna (IT)
  • Newcastle University (GB)
  • European Centre for Living Technology (IT)
  • University of Salento (IT)
Corresponding author
  • Fellermann, Harold
  • Stano, Pasquale
Venue
Frontiers in Bioengineering and Biotechnology
Published
2023-06-07
Volume
11
Pages
1210334-1210334
ISSN
2296-4185 (ISSN-L)
OpenAlex
W4379792915
Cited by
18
References
41
Citations by year
2026: 1 · 2025: 7 · 2024: 7 · 2023: 3
FWCI
1.60
Type
journal article
Language
en
Open access
Locations
  • Frontiers in Bioengineering and Biotechnology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Archivio istituzionale della ricerca (Alma Mater Studiorum Università di Bologna) (submittedVersion, cc-by)
  • AIRUS (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Newcastle University ePrints (Newcastle Univesity) (cc-by)
Funders
Engineering and Physical Sciences Research Council
Topics
  • Physical Sciences › Mathematics › Theoretical Computer Science › DNA and Biological Computing
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene Regulatory Network Analysis
  • Physical Sciences › Engineering › Biomedical Engineering › Molecular Communication and Nanonetworks
Keywords
artificial cells, recurrent neural networks, bacterial two-component systems, molecular memory, synthetic biology

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Recurrent neural networks in synthetic cells: a route to autonomous molecular agents? https://app.cochat.ai/papers/10.3389/fbioe.2023.1210334
The paper
Braccini, M., Collinson, E., Roli, A., Fellermann, H., & Stano, P. (2023). Recurrent neural networks in synthetic cells: a route to autonomous molecular agents? Frontiers in Bioengineering and Biotechnology. https://doi.org/10.3389/fbioe.2023.1210334

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.